The Epistemology of Memory: Evidentialism Meets Cognitive Psychology. Matthew Frise. Submitted in Partial Fulfillment of the

The Epistemology of Memory: Evidentialism Meets Cognitive Psychology by Matthew Frise Submitted in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy Supervised by Professor Earl Conee and Professor Richard Feldman Department of Philosophy Arts, Sciences and Engineering School of Arts and Sciences University of Rochester, Rochester, New York 2015

ii Dedication For Olivia, my bride

iii Biographical Sketch Matthew Frise was born in Los Gatos, California. He attended San José State University and was a President s Scholar and one of two Outstanding Graduating Seniors when he graduated summa cum laude with a Bachelor of Arts degree in Philosophy in 2007. He then attended the University of California, Santa Cruz, receiving a Regents Fellowship in 2007 and a Master of Arts degree in Philosophy in 2010. He began doctoral studies in Philosophy at the University of Rochester in 2010 and was awarded the University Fellowship. He received the Master of Arts degree in 2014, holding a Teaching Assistantship from 2011 to 2014. Under the direction of Professors Earl Conee and Richard Feldman he pursued his research in the epistemology of memory. The following publications were a result of work conducted during doctoral study: Epistemology of Memory, in The Internet Encyclopedia of Philosophy, edited by James Fieser and Bradley Dowden. (2015). Speaking Freely: On Free Will and the Epistemology of Testimony, Synthese 191 (7): 1587-1603, (2014). What God Only Knows: A Reply to Rob Lovering, Religious Studies 50 (2): 245-254, (2014).

iv The Mad, Bad, or God Argument Explained, Religious Studies 49 (4): 581 589, (2013). The Temporality Problem for Reliabilism [under review]. Internalism and the Problem of Stored Beliefs [under review]. Eliminating the Problem of Stored Beliefs [under review]. Preservationism in Epistemology [under review].

v Acknowledgments I thank all of my teachers, especially the Philosophy faculty at the University of Rochester, for shaping and educating me. I also thank my fellow graduate students for their friendship, support, and helpful philosophical discussion. A number of people deserve special acknowledgment a number so large that my extreme and general indebtedness to others is clear. For helpful conversation and comments on drafts of dissertation chapters, I thank Matthew Baddorf, Sven Bernecker, Caleb Cohoe, Earl Conee, Rich Feldman, David Henderson, Daniel Immerman, Ross Inman, Kevin McCain, Matthew McGrath, Kirk Michaelian, Bill Rowley, Tom Senor, Nico Silins, and Declan Smithies. Thanks to my friends and Epistemology Lunch-mates Matthew Baddorf, John Komdat, Kevin McCain, Dustin Olson, Bill Rowley, Jannai Shields, and Jarod Sickler, for talking so much epistemology with me. I am especially grateful to Bill and Kevin for their friendship, and for teaching me so much philosophy outside of the classroom. Some say that, in graduate school, one learns as much from one s peers as from one s professors. Bill and Kevin have helped make that true in my case. Amy Bray and Cheryl Bodensteiner have been anchors among administrative assistants. Thank you not just for providing incalculable help behind the scenes, but for helping with unparalleled efficiency. Many professors have been tremendously helpful. Thanks to Michael Tanenhaus for being on my committee, and to Scott Fraundorf and Kirk Michaelian for kindly pointing out

vi much psychological research relevant to my work. Thanks to Nico Silins for allowing me to take a fantastic epistemology course from him at Cornell University. Thanks to Ed Wierenga for letting me sit in on so many terrific classes, and for teaching the philosophy of religion with a perfect blend of analytic rigor, gentleness, and humor. Thanks to Ric Otte for introducing me in 2008 to the glories of analytic philosophy, for his guidance, friendship, and support, and for selflessly encouraging me transfer to the University of Rochester. My greatest academic debt is, and always will be, to Earl Conee and Rich Feldman. They do philosophy not just with world-class skill, but also with uncommon honesty and insight. I thank them for hosting an epistemology reading group or graduate epistemology course each semester, for their swift and frequent guidance, for their unfailingly good advice and feedback, and for being outstanding advisors. My parents, Duane and Sue, have been wonderfully encouraging throughout my graduate studies. They worked hard for decades at jobs they often did not love, so that I could go to college and work for decades at a job I love. Thank you. For their long love and wide support, thanks to my brother and sister-in-law, Nathan and Anna, to my in-laws Bobby, Denise, Robby, Bria, Bethany, and CJ, and to my friends Wayne Alder, Brian Ballard, Jed and Andrea Burke, Ethan and Colleen Hackett, and Bill and Rebecca Rowley. Finally, I thank my wife, Olivia, a beacon among brides. In order to support my studies and career she has run our entire home, even as I ve run it out of California to New York, to Missouri, and now to Texas. Without her, my time as a graduate student would have been far more solitary, poor, nasty, brutish, and long.

vii Abstract I use research from cognitive psychology to help reveal the nature of belief and of memory experience, and this clarity helps answer a fundamental question in epistemology: does one s justification for believing depend entirely on one s mental life? My two primary goals are to defend evidentialism, the most promising theory that answers yes to this question, and to challenge reliabilism, the most prominent theory that answers no. To these ends, I answer the problem of stored beliefs, a common objection to evidentialism that concerns memory. I argue that the cognitive psychology of memory reveals that, if either of the two leading theories of belief is correct, there is no problem of stored beliefs. I show that preservationism, the orthodox and unchallenged solution to the problem of stored beliefs, is either false or needlessly more complex than an evidentialist solution. Preservationism states that memory preserves the justification of preserved beliefs. Since reliabilism entails preservationism, it faces a serious problem. Finally, I argue that research on metacognition, the monitoring and controlling of cognitive processes, in memory favors evidentialism over reliabilism.

viii Contributors and Funding Sources Graduate study was supported by a University Fellowship and Teaching Assistantship from the University of Rochester, and by a visiting dissertation research fellowship from Saint Louis University and the John Templeton Foundation. (The opinions expressed here are those of the student and do not necessarily reflect those of the John Templeton Foundation). This work was supervised by a dissertation committee consisting of Professor Earl Conee (co-advisor) and Professor Richard Feldman (co-advisor) of the Department of Philosophy, and Professor Michael Tanenhaus of the Department of Brain and Cognitive Sciences. All work for the dissertation was completed independently by the student.

ix Table of Contents Introduction 1 Chapter 1 Internalism and the Problem of Stored Beliefs 4 Chapter 2 Eliminating the Problem of Stored Beliefs 31 Chapter 3 Preservationism in Epistemology 61 Chapter 4 Metacognition as Evidence for Evidentialism 101 References 134

1 Introduction Few epistemologists discuss memory in detail, yet most assume that the details support externalism over internalism about epistemic justification. I draw on work in the cognitive psychology of memory in order to achieve two main goals: to defend internalism, and to challenge a number of orthodox assumptions in the epistemology of memory. By internalism I mean the view that one s mental life alone determines what one is justified in believing. If internalism is right, then a belief can be justified for you if your experience supports it. The support would remain even if your experiences were systematically illusory. In contrast, externalism states that one s environment can affect what one is justified in believing even if it does not affect one s mental life. A main reason that philosophers think memory challenges internalism concerns stored beliefs. It seems that most of our justified beliefs beliefs about our past, about world history, about the meaning of words, etc. are usually stored in memory. But our thoughts and experiences cannot support more than a few of these beliefs at a time. Apparently, contrary to internalism, our mental lives are too impoverished to explain the justification of all these beliefs. Since internalism is a traditional and commonsensical view, its falsity would be unfortunate. However, cognitive psychological research, I claim, tells an internalist-friendly story about our mental lives. In Chapter 1 I clear space for sharing this story by challenging the two leading internalist attempts to solve this problem of stored beliefs. First I dispute the attempt from epistemic conservatism, which states that believing justifies retaining belief. Then I defend

2 the attempt from dispositionalism, which assigns a justifying role to dispositions, from some key objections. But by drawing on cognitive psychological research I show that, for internalism, there is not yet reason to suppose dispositionalism solves the problem of stored beliefs. In Chapter 2 defend a disjunctive solution to the problem of stored beliefs for internalism. Does having a belief that p require having a special relation to a mental representation that p? On the leading No answer, a belief is roughly a set of dispositions. I argue that dispositionalism about justification follows. Since these dispositions are mental, internalism has the resources for counting the right stored beliefs as justified. If the answer is Yes, then there are no stored beliefs at all. Extensive research in cognitive psychology suggests that memory alters information considerably during several stages of processing, and that the relation we bear to the information memory stores is underdetermined. As a result, I argue, memory doesn t store representations that match the content of our alleged stored beliefs, and we don t bear the special relation to the representations that memory does store. So if the answer is Yes, then there are no stored beliefs, and there is no problem of stored beliefs. We still have stored beliefs in a derivative sense that I explicate in detail, and it is in a derivative sense that they are justified. In Chapter 3 I offer three dilemmas for preservationism, an orthodox and previously unchallenged view in the epistemology of memory. Preservationism states that memory preserves any justification of a preserved belief. More precisely: if S formed a justified belief that p at t 1 and retains in memory a belief that p until t 2, then S s belief that p is prima facie justified via memory at t 2. My belief that Plato taught Aristotle is justified now because it was

3 justified when it was formed and because it has been retained in memory ever since. Each dilemma shows that preservationism accommodates our intuitive judgments about justification only if it objectionably fails to be parsimonious. Externalists endorse no other view about what sustains justification as time passes and as beliefs become stored. Since several theories of justification entail preservationism, my dilemmas challenge these theories as well. In Chapter 4 I challenge the common assumption that externalism alone allows for a naturalized epistemology, i.e., an epistemology that is informed by scientific research and methodology. Notably, all epistemological discussion of research on metacognition the monitoring and controlling of cognitive processes assumes an externalist framework. I argue that research on metacognition in memory supports an internalist form of evidentialism over a popular externalist theory: reliabilism. Evidentialism states that the justified attitude for a subject toward a proposition is the attitude that fits the subject s evidence. I argue that evidentialism uniquely predicts that, in an ordinary case of recollection, the subject s evidence fits her justified beliefs. Research on metacognition shows that certain types of mental states accompany ordinary recollection, and I argue that these states are fitting evidence; i.e., evidentialism s prediction is accurate. Reliabilism does not share this prediction. Moreover, it does not both accurately and uniquely predict anything in ordinary recollection cases. So, I argue, these cases better support evidentialism. Support for my argument reveals that reliabilism is importantly unverifiable.

4 Chapter 1: Internalism and the Problem of Stored Beliefs 1. Introduction Most justified beliefs, it seems, are stored in memory. At any given time, our justified beliefs about our past, geography, philosophy, world history, and most other topics are nonoccurrent. We are not thinking about, reasoning from, acting from, or having an experience concerning these beliefs or their content. Of course, a belief can be both stored and occurrent, just as a song can be both stored and playing on one s computer. But most justified beliefs are merely stored (hereafter stored). They are in no way before our minds. The problem of stored beliefs is the problem of satisfactorily explaining how the stored beliefs which seem justified are indeed justified. In this chapter, I argue that internalists about epistemic justification have not solved this problem for their view. Internalism, at a minimum, states that possible beings who are mentally identical are, to the same degree, justified in believing the same propositions. 1 No contingent fact about a subject s environment affects her justification unless it affects her mental states. Someone whose experiences are identical to yours but who is massively deceived by an evil demon is justified in believing the same propositions you are. Internalism also states that all justifying features are mental. It is controversial among internalists exactly what these features are, and controversial whether these features must be specially accessible to the subject. This chapter remains neutral on these controversies. 1 Cf. Conee and Feldman (2004: 56).

5 Some philosophers deny internalism because they think it cannot solve the problem of stored beliefs. 2 They think that even if internalism accounts for the justification of some justified stored beliefs, our mental lives are too impoverished to account for the justification of all. If internalism cannot solve the problem, it is significantly discredited. What s more, one of the best arguments for internalism would fail. The argument alleges that internalism s implications match our intuitions about when a subject is or is not justified in believing a proposition. 3 Internalism gets the cases right. If internalism cannot solve the problem of stored beliefs, it gets some cases wrong. While I think internalism can solve this problem, my aim in this chapter is to challenge the two leading internalist attempts. 4 First I dispute the attempt from epistemic conservatism. Then I examine the attempt from a view I call dispositionalism, which assigns a justifying role to dispositions. I clarify and develop dispositionalism, and defend its attempt from some key objections. But, by drawing on cognitive psychological research, I show that the problem of stored beliefs remains. 2 See Goldman (1999; 2009; 2011), Senor (1993; 2009), and Williamson (2007). Cf. Harman (1986) and Huemer (1999). Pappas (1980) discovered the problem, Senor (1993) directed it at internalism, and Goldman (1999) named it. 3 Cf. Conee and Feldman (2004: 58-61). 4 I defend an internalist solution in Chapter 2. It is worth noting that externalism faces the problem of stored beliefs too. According to externalism, a contingent feature of a subject s environment can affect her justification without affecting her mental life. But defenders of externalism think they can readily solve the problem. I challenge their attempt in Chapter 3.

6 Two preliminary remarks are in order. First, the problem of stored beliefs is an importantly infelicitous name. It seems there is in fact more than relevant one problem, or that there are underappreciated dimensions to the problem of stored beliefs. As stated so far, the problem concerns only stored doxastic justification, that is, the justification of a stored doxastic attitude one has. But there are stored beliefs which some subjects should, but do not, have. And there are stored beliefs a subject should have and indeed has, but has for the wrong reasons. How does a theory of justification satisfactorily explain these should-claims? In other words, it seems subjects have stored propositional justification, justification for believing a proposition which the subject does not believe or does not believe justifiedly, where this proposition is not before the subject s mind. Consider: DESI Desi has a stored belief that fairies exist, and she has good enough reason to believe that they do: she remembers that her clearly trustworthy brother told her so. But what caused and currently sustains her belief that fairies exist is simply her desire that they exist. She has no relevant higher-order belief about how she has good reason to believe that fairies exist. Because of its basis, Desi s stored belief that fairies exist is not justified (her belief is improperly caused, she has no suitable higher-order belief about it, or whatever). But Desi has justification for, or is justified in, believing that fairies exist. She would have this justification even if she did not believe that fairies exist. A theory of justification that does not account for this has not solved the problem of stored beliefs, even if accounts for the

7 justification of all of Desi s justified stored beliefs. 5 There is a propositional problem of stored beliefs in addition to the more familiar doxastic problem of stored beliefs. The significance of this distinction will become clear in the next section, although for simplicity I often just discuss the more familiar problem. Second, an adequate solution must satisfy the Constraint. The Constraint, commonsensically, states that no subject or attitude has a justificatory status that it lacks. X satisfies the Constraint iff X does not count any possible subject S (or doxastic attitude of S s) as having a justificatory status that S (or her attitude) lacks. Obviously, an adequate solution must not solve either problem of stored beliefs in a way that attributes too much justification. It must not, for example, count stored beliefs that seem unjustified as justified, or count a subject as having justification for believing a proposition when her suspending judgment or disbelieving instead seems justified. Overly liberal solutions will not do. 2. Epistemic Conservatism Epistemic conservatism appears to solve the problem for internalism. According to Matthew McGrath (2007: 14), epistemic conservatism is the view that if one believes that p, then one is prima facie rational in retaining that belief. 6 One is justified in continuing to 5 The problem of stored beliefs is not fully apt in other ways. There is a problem if there is a single stored belief the justification of which is not explained. So, beliefs is misleading. And suspended judgment and disbelief can also be stored bearers of justification. To focus on the problem concerning just beliefs is not to exhaust the relevant problems. For simplicity, this chapter sets these observations aside. 6 Cf. Harman (1986: Ch.4) and McCain (2008: 186). In light of an objection, McGrath (2007: 21) eventually restates the view in a restricted form: If S believes that p, then S is rational to retain belief that p iff condition C

8 believe a proposition as long as one lacks reason to abandon belief. If Maria believes that dairy milk comes from cows, she is overall justified in retaining that belief if she has no defeaters no reason to think that dairy milk does not come from cows, that her memory is especially defective, that she originally formed her belief for no good reason, and so on. Her having this justification does not require her to have, or to be disposed to have, an experience with memorial phenomenology that represents memory as generally accurate or represents dairy milk as coming from cows. And it does not matter whether Maria s belief is occurrent or stored. It still justifies her in retaining the belief. Epistemic conservatism is typically understood to be a form of internalism. It suggests that beliefs justify, and beliefs are mental states, so it suggests that mental states justify. But epistemic conservatism states only a sufficient condition for having justification, not a necessary condition. For all that epistemic conservatives have said, non-mental differences could result in justificatory differences. Still, epistemic conservatism is compatible with internalism, so it could provide internalism with a solution to the problem obtains. McGrath (2007: 21 n.39) expresses doubts about this formulation, and does not attempt to defend any particular interpretation of C. In the absence of such an interpretation, the epistemic conservative solution to the problem of stored beliefs faces the first objection I will give above. And even in the presence of such an interpretation, the epistemic conservative solution still faces my other objections. Epistemic conservatism may not be an ideal name for the view discussed in this section, since it sometimes refers to a genus of views, whose species include doxastic conservatism and phenomenal conservatism. Doxastic conservatism is the view that McGrath and others call epistemic conservatism, and phenomenal conservatism is the view that if it seems to S that p, then S has prima facie justification for believing that p. Since I am primarily engaging McGrath here, I follow his terminology.

9 of stored beliefs. And McGrath (2007: 21-2) thinks epistemic conservatism solves the problem. The stored beliefs that we are inclined to count as justified are typically ones that we do not have overall reason to abandon. So, we are justified in retaining them. Epistemic conservatism seems to solve the problem, and no one has objected to it as a solution. I will give three reasons for thinking that epistemic conservatism, even if true, does not solve the problem. The first reason will be clearer when it is clearer what epistemic conservatism states. It states a sufficient condition for justifiedly retaining a belief. Retains picks out a dynamic relation. It identifies a relation between objects over time. Epistemic conservatism states that a subject s having a belief at one time prima facie justifies the subject in having it a moment later. If the subject has the belief a moment later, it is justified. Epistemic conservatism is a theory of diachronic justification. More precisely: EC1. If S believes that p at t 1, then S is prima facie justified in believing that p at t 2. EC2. If S believes that p at t 1 and at t 2, then S s belief at t 2 is prima facie justified. EC1 states epistemic conservatism s sufficient condition for propositional justification (for simplicity I often drop the prima facie ). EC2 states epistemic conservatism s sufficient condition for doxastic justification. McGrath and other epistemic conservatives do not explicitly state their view as EC1 and EC2. But it is hard to see what other theories of propositional and doxastic justification could be picked out by the temporal relation of rational in retaining. (And a theory of doxastic justification, not just propositional justification, appears to be picked out, since what the subject allegedly is rational in retaining is a belief.) Epistemic conservatism is not a

10 theory of synchronic justification, a theory about what at a time justifies belief at that time. This is for the best, because epistemic conservatism as a theory of synchronic justification, a theory about only t 1, seems false: EC3. If S believes that p at t 1, then S is prima facie justified in believing that p at t 1. EC4. If S believes that p at t 1, then S s belief at t 1 is prima facie justified. EC3 seems false because a belief does not seem to be self-supporting at a time. In fact, McGrath (2013: 245) agrees. Having a belief does not by itself support its content, although having had belief might support its content it is plausible that we should trust our past selves, other things being equal. And even if a belief could support itself, EC4 seems false. Arguably, in order for a belief to be justified, it must be caused by what justifies it. 7 But a belief cannot cause itself at a time, just as a fire cannot cause itself at a time. At best, a belief that p at one time can cause a belief that p a moment later, just as a fire at one time can cause a fire a moment later. So epistemic conservatism is best understood as stating EC1 and EC2, but not EC3 and EC4. A first worry for epistemic conservatism as a solution to the problem of stored beliefs is that it seems to get DESI wrong. EC2 implies that Desi s stored belief that fairies exist is prima facie justified, since Desi had that belief a moment ago. The bad basis of the belief does not prevent it from being prima facie justified. Unless Desi has some defeater for 7 A theory placing a causal requirement on doxastic justification can still be internalist, if it states that mental features alone propositionally justify. For such an internalist theory, see McCain (2014: Chap. 5, 6).

11 her belief, it is according to epistemic conservatism justified overall. But her belief is not justified overall. The epistemic conservative could reply that there is a defeater for Desi s belief. McGrath (2007: 18) is fairly liberal about defeating conditions: defeaters, possessed or not, which are constructible from one s current perspective, are potential defeating conditions. But even if we grant this view about defeating conditions, it is implausible that there must be a defeater constructible from Desi s perspective. It is rarely clear from one s current perspective what originally caused or now sustains any of one s beliefs, even when one s good reasons caused and sustain one s beliefs. We can stipulate that Desi cannot tell that her fairy belief is based on desire, and so she lacks the materials to construct a defeater. Consequently, epistemic conservatism counts her unjustified stored belief that fairies exist as justified full stop. Epistemic conservatism does not satisfy the Constraint. 8 Second, while EC2 can account for the justification of a subject s belief that p at t 2, it does not thereby account for the justification of the subject s belief that p at t 1. This is a problem because not all possible justified stored beliefs are believed at some earlier time. Consider: 8 McGrath (2013: 245) independently notes that cases like DESI threaten epistemic conservatism. He thinks they are potential counterexamples to his theory. (For that reason, in personal correspondence, he found EC1 much more attractive than EC2). It is worth emphasizing that even if they are not counterexamples, they show that epistemic conservatism violates the Constraint and therefore fails to solve the problem of stored beliefs. It is also worth noting that epistemic conservatism and dispositionalism (see section 3) are silent on how it is that stored suspended judgment can be justified. These views need to be expanded in order to account for this justification.

12 VIC Vic is a victim in a memorial skeptical scenario. An evil demon popped Vic into existence at noon today he did not exist until then. At noon Vic s mental states are identical to Ric s. Ric came about in the ordinary way, long before noon. He believes many things and has many apparent memories. Vic, being Ric s mental twin, believes the same things and has the same apparent memories. Many of these beliefs are justified and stored. There is no reason to doubt that VIC is possible. Victims in memorial skeptical scenarios seem able to have stored beliefs when they come into existence, and it seems that many of these beliefs could be justified, especially when the victims have mental twins with many justified stored beliefs. What s more, internalism is often motivated by the new evil demon problem, which centers on a structurally parallel case to VIC but which concerns externalworld skepticism, not memory skepticism. Internalists will have special reason to grant that VIC is possible. VIC, if possible, reveals that epistemic conservatism poorly solves the problem of stored beliefs. Since Vic does not exist before noon, he of course does not believe anything prior to noon. So, Vic has no antecedent beliefs which could explain the justification of any of his stored beliefs at noon. McGrath mistakenly supposes that for every justified stored belief that p, there is a prior belief that p. But a belief, justified or not, can be stored even if it has not been previously held. A belief is stored so long as it is not before the subject s mind. EC2, even if true, does not solve the problem of stored beliefs. To be clear, the basic reason is this. Epistemic conservatism is just a thesis about justified belief retention, not about justified belief formation. It does not account for the justification of any belief that is stored directly at the time of its formation.

13 One might reply as follows. If a subject s belief that p is justified, then the belief is caused by that which justifies the subject in believing p. And causation always takes time. Whatever it was that brought about Vic s beliefs at noon, it was not something that justified him in believing p a moment before. Since Vic did not exist until noon, prior to noon nothing justified him in believing anything. Vic s beliefs at noon are therefore not caused by their justifiers, and so they are not doxastically justified at noon. So, VIC does not reveal that there are justified stored beliefs which epistemic conservatism cannot account for. I will grant that causation always take time, although this thesis does face objections. VIC still introduces trouble for epistemic conservatism as a solution to the problem of stored beliefs. For EC1 does not explain Vic s propositional justification at noon. We can grant that Vic has no justified stored beliefs at noon, since none of his beliefs are at noon based on his justification. But it is plausible that at noon Vic is justified in believing many things, including the contents of some of his stored beliefs. Yet EC1 does not credit Vic as having justification for believing anything he believes. The antecedent of EC1 is not satisfied, since Vic did not exist a moment before noon. Epistemic conservatism fails to solve the propositional problem of stored beliefs. This may tempt us to understand epistemic conservatism as stating EC3 after all. According to EC3, at any time, S s believing that p prima facie justifies S in believing that p. S always has prima facie justification for anything she believes. Because Vic has stored beliefs at noon, he would at noon have justification for believing the contents of all his stored beliefs. Above I noted that EC3 does not capture the temporal character of retains and that EC3 seems false, but let s set these details aside. EC3 still does not ultimately help with the

14 propositional problem of stored beliefs presented by VIC. Suppose Ric at noon fails to believe some proposition q that he has justification for believing and that is not before his mind. Since Vic is Ric s mental twin, internalism entails that he too is justified in believing that q at noon. EC3 fails to account for either subject s justification for believing q by stipulation, at noon neither subject has a belief that q which could provide this justification. Yet accounting for this justification is part of the propositional problem of stored beliefs. So EC3, even if true, does not solve the problem. Here is my final objection to epistemic conservatism as a solution to the problem of stored beliefs. As we have seen, epistemic conservatism (understood as just EC1 & EC2) does not count Vic as having justification for believing the contents of any of his stored beliefs at noon. If epistemic conservatism is a complete solution to the problem of stored beliefs if it tells the full story of the justification of our stored beliefs then Vic is unjustified in believing the contents of all of his stored beliefs at noon. But according to EC1, a moment after noon, Vic is prima facie justified in believing the contents of all stored beliefs he had at noon. We can stipulate that none of Vic s mental states, conditions, or events change between noon and a moment after noon. We can even stipulate that there is no intra-mental causal connection between Vic s mental states at these times; Vic s mental life a moment after noon is not caused by his mental life at noon, but rather is caused by whatever brought about his mental life at noon. A moment after noon, his noon beliefs give him justification he lacked at noon. So, Vic is mentally identical to himself at these two times, but his justification is not identical. Thus epistemic conservatism, if we do not supplement it with some account of Vic s noon justification, unexpectedly denies internalism.

15 Consequently, it would not solve the problem of stored beliefs for internalism! And, of course, if we do supplement epistemic conservatism with an account, then it is not by itself a solution to the problem of stored beliefs. Contra McGrath, even if epistemic conservatism is true, it does not solve the problem of stored beliefs. The three reasons for believing this may even refute epistemic conservatism. I will not argue here that they do. But they do undermine epistemic conservatism considerably; one of its main alleged assets is that it accounts for our justification that concerns memory. 9 3. Dispositionalism 3.1 Dispositionalism Developed and Defended Several internalists have defended some version of a view that I will call dispositionalism: If S has disposition X, then S is prima facie justified in believing that p. Dispositionalism is a schema. Different versions of it assign different interpretations to X. On all versions, a disposition of a certain sort justifies believing a related proposition. 9 McCain (2008), McGrath (2007), and Poston (forthcoming) claim that epistemic conservatism has this asset.

16 Dispositionalists have tried to solve the problem of stored beliefs. 10 A disposition pertaining to memory can justify a stored belief, they claim, even when the disposition is not manifest. In this section I will defend and develop the most recent and promising dispositionalist attempt to solve to the problem namely, Earl Conee and Richard Feldman s (2011: 303-5) attempt. In the next section, I show that it is unsuccessful. Conee and Feldman fill out dispositionalism as follows: DR. If S has a disposition to recollect p, then S is prima facie justified in believing that p. 11 A disposition to recollect p is a disposition to recall p as known, not simply as believed. Having this disposition does not require p to be true, much less known. If the disposed recollection is strong and clear enough, and undefeated, it justifies strongly enough for knowledge. Conee and Feldman (2011: 304-5) suggest that a subject has a recollective disposition toward p when she has learned and has not forgotten that p, where learned is non-factive. That is: 10 See Robert Audi (1995: 37), Conee and Feldman (2011: 303-5), Carl Ginet (1975: 154-7) and Tommaso Piazza (2009). Cf. Conee and Feldman (2004: 236-7). For related attempts, see Conee and Feldman (2004: 67-9) and McCain (2014: Chap. 3). 11 Conee and Feldman describe recollective dispositions as defeasible, justifying evidence. For simplicity I omit talk of evidence, and I mostly talk about prima facie justification rather than defeasible justifiers. Nothing hangs on this.

17 LNF. If S has learned and not forgotten p, then S has a disposition to recollect p. But Conee and Feldman allow that recollective dispositions can come about via brain malfunction or tampering, and that the disposition still defeasibly justifies, whatever its historical origins. So it is best to understand learned broadly. I will call the conjunction of DR, LNF, and Conee and Feldman s account of a recollective disposition Conee and Feldman s dispositionalism. How might their dispositionalism solve the problem of stored beliefs? They (2011: 304) suggest that ordinarily a subject with a justified stored belief that q has the potential to bring q to mind with the phenomenology of activating a memory, specifically, the memory that q. The subject has a disposition to recollect q. Conee and Feldman s dispositionalism seems able to account for the propositional and doxastic justification relevant to the problem of stored beliefs. Thomas Senor (2009) and McGrath (2007) apparently object to any form of dispositionalism as a solution to the problem of stored beliefs. 12 But Conee and Feldman s dispositionalism, I argue, is safe from the objections. Senor (2009) says, regarding the problem of stored beliefs: This problem might be thought resolved by appealing to counterfactuals or dispositions. That is, a stored belief might be thought to be prima facie justified iff were this belief to become occurrent, it would be accompanied by the appropriate memory image or seeming. Yet this condition is pretty clearly false. The phenomenology of recollection depends crucially on the context in which the belief 12 Goldman (1999) does as well, but his objections only threaten a form of dispositionalism that Conee and Feldman (2011) do not endorse. Since Senor and McGrath identify no particular form of dispositionalism as their target, their objections seem intended for all forms.

is recalled. Typically, the experience one has when one recalls a belief depends at least on how the recollection was cued, how much attention one is currently paying to the recalled belief, and what other beliefs are then occurrent. 18 The objection to the dispositionalist solution seems to be that DR is false: a recollective disposition does not justify. Senor assumes that a disposition is analyzable in terms of a counterfactual conditional that involves a particular stimulus condition. The relevant stimulus condition is the justified stored belief s becoming occurrent. Senor supports his objection by denying that, in this condition, the subject would have a justifying memory experience. This is because the phenomenology of actual recollection is importantly sensitive to contextual factors. In some contexts, the subject would have justifying phenomenology, but not in others. Since mere recollection does not justify, a mere recollective disposition does not either. If Senor s objection is correct, Conee and Feldman s dispositionalism does not solve the problem of stored beliefs. It is worth noting that the defender of Conee and Feldman s dispositionalism need not grant Senor s conditional analysis of dispositions. It is difficult to state well what a disposition is, and the conditional analysis is at best highly controversial. We are free to pick an alternative analysis or to suppose that there simply is no analysis. We can understand a disposition as we do pre-analytically, and think of a disposition to recollect p as similar to one s values, or moods, or character traits, like shyness or irritability or generosity. It is much easier to recognize that someone values honesty than it is to find a true counterfactual that corresponds to her valuing honesty. And, even if we accept a conditional analysis, we are free to identify a much broader stimulus condition than Senor does.

19 Still, even if we grant Senor s assumptions, the support for his objection is inadequate. His support is that a subject s circumstances crucially determine the phenomenology of her recollective experience. The what it s like aspect of memory experience is importantly context-sensitive. But phenomenology that is importantly contextsensitive need not relevantly vary. And it is not clear that recollective phenomenology relevantly varies. True, it varies in strength and in degree of familiarity. In some contexts the memorial phenomenology a subject has while recollecting p is more vivid, and in other contexts less. But in all these contexts the same content is presented to the subject with some memorial phenomenology the proposition is brought to mind with, as Conee and Feldman put it, the phenomenology of activating a memory. Even though the phenomenology of actual recollection is context-sensitive, it appears sufficiently invariant to justify invariantly. So there is no reason to deny that the recollective disposition justifies. Senor s objection may be correct, but it lacks support. Conee and Feldman s dispositionalism may very well solve the problem of stored beliefs. McGrath (2007: 11) resists dispositionalist solutions for a different reason. He claims that One might have the disposition [to enjoy memorial phenomenology] without having any [justifying] evidence at all. He tries to illustrate this with an example like the following: FARGO Fargo has forgotten his childhood home phone number. But he is disposed, when asked precisely about it, to undergo memorial phenomenology regarding it. If asked

Is it true that your childhood home phone number was xxx?, Fargo would recall that it was xxx. 13 20 McGrath concludes that In such a situation, before being asked, one is not rational to believe that p. And so one s being disposed must not be a way of having the [justifying] evidence. His point seems to be that dispositionalism would yield too much justification. One could in very particular circumstances be disposed to recollect that p, and therefore, unintuitively, count in all circumstances as justified in believing that p. In effect, McGrath is claiming that a dispositionalism like Conee and Feldman s violates the Constraint. However, FARGO reveals no flaw in a dispositionalist solution. FARGO stipulates that Fargo has forgotten that his childhood home phone number (hereafter his number) was xxx. On Conee and Feldman s dispositionalism, the sufficient condition for having a recollective disposition is: learning and not forgetting. Their dispositionalism does not imply that Fargo has a disposition to recollect that his number was xxx. So, their dispositionalism does not attribute to Fargo justification for believing that his number was xxx. Similarly, FARGO stipulates that Fargo is disposed to recollect that his number was xxx when he is asked about it. It does not follow that he has a disposition to recollect that his number was xxx. If S is disposed to do A in some circumstances C, it does not follow that S is disposed to do A simpliciter. There is no reason to suppose that Fargo, prior to being 13 The details of FARGO are from McGrath (2007: 11). I ve named the case and its subject. McGrath uses FARGO to present an objection that he says is Goldman s (1999: 278-9). But FARGO improves the case Goldman uses. In Goldman s, the subject s relevant disposition involves information that the subject does not in any way already possess, so it is doubtful that the subject has a disposition to recollect that information. So I discuss only FARGO.

21 asked about his number, has a disposition to recollect it. So there is no reason to suppose that on dispositionalism he, prior to being asked about his number, is justified in believing it was xxx. Further, once Fargo is asked, it is not implausible that he is justified in believing it was xxx. There is no reason here to think that Conee and Feldman s dispositionalism violates the Constraint. McGrath s objection is unsuccessful. I now offer two improvements for Conee and Feldman s dispositionalism that better position it to solve the problem of stored beliefs. Each improvement calls for a broader understanding of recollective disposition. First, Conee and Feldman identify a recollective disposition that p as a disposition to recall p as known. To recall p as known requires one to employ the concept of knowledge. This may demand too much sophistication from many ordinary subjects in ordinary cases, such that there are too few recollective dispositions to solve the problem of stored beliefs. Fortunately, being disposed to recall p as true also seems to justify believing that p. Nothing is lost and something is gained if we count such a disposition as recollective. Conee and Feldman s dispositionalism seems more promising on this broader understanding of recollective disposition. Second, in some cases where p is not before the subject s mind, it seems the subject has justification for believing p yet lacks a disposition to recall p as true or known. Philosophers and psychologists of memory standardly endorse a tripartite distinction regarding types of memory: semantic (memory of a proposition), episodic (memory of an

22 experienced event), and procedural (memory of skill or of how to perform an action). 14 On Conee and Feldman s dispositionalism, dispositions rooted in semantic memory justify all that is not before the subject s mind but that is nonetheless justified. But if certain dispositions justify, it is plausible that some, perhaps many, are rooted in episodic memory. Episodic memory enables a re-experiencing of events as experienced before. Being disposed to recall a set of experienced events that are suitably related to p justifies believing p, even in the absence of a disposition to recall p itself. Consider: BUSTER Buster is justified in believing that he took the bus today. But he isn t disposed to recall that as true. Rather, his episodic memory disposes him to have experiences that justify his believing it. Buster is inclined to activate with memorial phenomenology first-person experiences of his bus ride memorial imagery of passengers seated nearby, of trees whizzing in the windows; a memorial olfactory experience of the smoker beside him; a memorial auditory experience of the bus droning beneath him. He is inclined to feel these experiences as connected in a certain temporal order with other experiences. Buster is not disposed to recall that he took the bus today. He is instead disposed to recall many experiences that support believing that proposition. Now, one might think that, unless Buster consciously associates those experiences with the proposition, the (disposition to have the) experiences cannot justify him in believing the proposition. But this seems to over- 14 Philosophers and psychologists use various terms to make more or less the same distinction. In psychology, Tulving (1972) popularized the semantic/episodic memory distinction. In contemporary philosophy, Russell (1921/1995) and Martin and Deutscher (1966) influentially discuss episodic memory, and Locke talks about it even earlier.

23 intellectualize things. When a subject does not consciously associate her recalling p with p, her recalling still justifies believing that p. It is hard to see why it would be different when the subject recalls experiences that support p. Conee and Feldman s dispositionalism is more promising if, on it, dispositions to recall events as experienced by the subject count as recollective, and therefore as justifying. Cases like BUSTER seem common, and a solution to the propositional problem of stored beliefs must account for them. Similarly, semantic memory seems to justify some stored beliefs, but not always by disposing us to recall the content of these beliefs as true or known. Consider: GIL Gil has a stored belief that (i) he swam in the Gulf of Mexico. He is disposed to recall as true that (ii) he swam at a beach in Pensacola, Florida, and that (iii) Pensacola beaches are on the Gulf of Mexico. But he is not disposed to recall as true that he swam in the Gulf of Mexico. Rather, he is disposed to infer it automatically upon recalling (ii) and (iii). It seems Gil s stored belief is justified by some of his dispositions. By recalling (ii) and (iii), it will become clear to Gil in a memorial way that he swam in the Gulf of Mexico. A disposition to recall a set of propositions other than p seems able to justify believing that p. One might object that Gil must think about (ii) and (iii), or consciously associate them with (i), in order for his inference to justify believing (i). But, again, this seems to overintellectualize matters. In sum, Conee and Feldman s dispositionalism is more promising if we understand recollective disposition more broadly. Each of the following should count as a recollective disposition toward p: (a) a disposition to recall p as known, (b) a disposition to recall p as

true, (c) a disposition to recall experienced events that jointly and clearly indicate that p, and (d) a disposition to recall a set of propositions that jointly and clearly indicate that p. 24 3.2 Dispositionalism Indisposed For two reasons Conee and Feldman s dispositionalism still does not solve the problem of stored beliefs, and it is not clear that it can be improved further. First, Conee and Feldman s dispositionalism runs afoul of the Constraint. Recall that LNF states a sufficient condition for S s having a recollective disposition toward p: S s having learned and not forgotten p. However, there is considerable psychological research suggesting that we forget far less than we are inclined to think we do. In fact, very little that we learn and store in longterm memory is ever forgotten. 15 Rather, it is just not always directly accessible, and this can mislead us into thinking it is totally inaccessible. We retain the information and can retrieve it if cued properly. If this psychological data is roughly accurate, then Conee and Feldman s dispositionalism fails to satisfy the Constraint. A case illustrates this: MARGO Margo learned and has not forgotten her childhood home phone number. If asked Is it true that your childhood home phone number was xxx?, Margo would recall that it was xxx. But if merely asked What was your childhood home phone number?, Margo would not recall it. And if merely asked Can you recall your childhood home phone number?, Margo would not recall it. In fact, she is disposed to seem to herself unable to recall the number. 15 Koutstaal and Schacter (1997) review the psychological research on this. Cf. Wagenaar (1986). Bjork and Vanhuele (1992: 156) go so far as to claim that any information that makes it into long-term memory remains there essentially forever. For philosophical discussion, see Michaelian (2011b: 403-4).